repo owner: Danni (Danqing) Zhang ([email protected])
main contributors: Balaji Rama ([email protected]), Shiying He ([email protected]) and Jingyi Ni ([email protected])
In our exploration of building multi-agent systems, we have investigated AutoGen, CrewAI, LangGraph, and MetaGPT, but failed to find a multi-agent system that intuitively makes multi-agent systems more efficient. We have found that most examples provided by these frameworks could be accomplished with just one agent. We reimplemented the multi-agent examples using only one agent with a set of self-defined tools. However, we then realized that this system is not scalable when we have more and more tools. But if we can categorize the agents with different sets of tools into categories, then we are building a hierarchy of agents, where we could accomplish more types of tasks. At the same time, because of this design, the execution of sub-agents is naturally parallelized by parallel function calling, since agents use sub-agents as tools.
First set up virtual environment
$ python3 -m venv venv
$ . venv/bin/activate
$ pip install -r requirements.txtThen set up a Supabase project to retrieve the API keys, and create a table if you want to save the results to a Supabase table. Please add the below columns to your Supabase table, and disable RLS.
| Column Name | Data Type |
|---|---|
| agent | text |
| depth | int8 |
| response | text |
| role | text |
| prompt_tokens | int8 |
| completion_tokens | int8 |
| input_cost | float8 |
| output_cost | float8 |
| total_cost | float8 |
| model_name | text |
| meta_task_id | text |
| task_id | int8 |
Then please create a .env file, and update your API keys:
cp .env.example .envAfter configuring your API keys, you can run main.py, where the sub-agents are already parallelized through parallel function calling.
cd litemultiagent
python demo_agent.pyThen we can get the total execution time for all tasks: 67.07 seconds.
You can parallelize the execution of the tasks of main agents:
python demo_agent.py --main-workers-parallelThen it's further sped up. Total execution time for all tasks: 29.34 seconds.
If we neither the tasks at main agent level or sub agent level are paralleled, the total execution time for all tasks: 88.43 seconds
For how to contribute, see CONTRIBUTE. If you would like to contribute to the codebase, issues or pull requests are always welcome!